lego573402/bible-vocabulary-difficulty

Bible vocabulary-difficulty metrics, 12 translations

Per-verse reading-difficulty metrics plus a cross-language book-name table keyed on USFM codes. Produced by bible-reader — see scripts/export_dataset.py.

No verse text

This dataset contains references and derived metrics only, never verse text. That is deliberate: it keeps translations under copyright (NASB) publishable as derived data, and it keeps the download small. Fetch the texts themselves from their own sources.

Metrics

column	meaning
ref	verse reference as it appears in the source text
book_usfm	USFM book code — the join key across translations
chapter, verse	parsed from ref
total_count	tokens in the verse after language-aware tokenization
difficulty_rank	the useful one, see below

difficulty_rank is the size of the vocabulary you need to read the verse. Rank every distinct word form by frequency within its own translation (1 = most frequent), then report the smallest N such that knowing the top N forms covers ≥95% of the verse. Sorting a translation by difficulty_rank ascending gives a learning order. It is computed per translation, so ranks are comparable within a translation but not across them.

Tokenization is language-aware: Snowball stemming for the 15 supported European languages, niqqud/cantillation stripping for Hebrew, diacritic stripping for Greek, harakat stripping for Arabic.

Note what is not here: comprehension rate and known-word counts. Those are scored against one reader's vocabulary, so they describe a person rather than a text. Everything published here is vocabulary-independent.

Translations

config	name	language	verses
nasb	NASB (English)	en	31,102
wlc	Biblical Hebrew OT (WLC)	he	23,213
gnt	Biblical Greek NT (Byzantine)	el	7,953
modern-he-nt	Modern Hebrew NT (Delitzsch)	he	7,956
valera	Reina Valera (Spanish)	es	31,084
luther1545	Luther 1545 (German)	de	31,098
ls1910	Louis Segond 1910 (French)	fr	31,170
synodal	Synodal (Russian)	ru	37,098
almeida	Almeida (Portuguese)	pt	31,101
riveduta	Riveduta (Italian)	it	31,102
statenvertaling	Statenvertaling (Dutch)	nl	31,312
arabicsv	Smith & Van Dyke (Arabic)	ar	31,102

books.csv

One row per USFM book code. There are two kinds of name column and the difference matters:

• english, hebrew, greek are curated canonical names — what the book is actually called in that language.
• The per-translation columns (nasb, wlc, synodal, …) hold the book string exactly as it appears in that text's references, so they join against the ref field in the Parquet files. For wlc and gnt those are the Latin abbreviations the source files ship (Gen, Matt) — the real names are in hebrew and greek.

Canons differ — Statenvertaling carries Baruch, the Russian Synodal text carries the full Orthodox deuterocanon and orders the NT with the Catholic epistles before the Pauline ones — so a blank cell means the book is absent from that translation, not that data is missing.

Notes on the Hebrew and Greek columns:

• Hebrew names are unvocalized, which is the standard citation form. Hand-transcribed niqqud is a reliable source of silent errors and wrong pointing is worse than none. OT names are the Tanakh titles; NT names are the Delitzsch Hebrew NT's, in short form. Of the deuterocanon only Sirach has a Hebrew original, so the rest are blank.
• Greek is polytonic. NT titles are the traditional ones and match the Byzantine text here. OT titles are Septuagint names, included as reference data even though this corpus has no Greek OT — note that Samuel and Kings are the four books of Βασιλειῶν, and that LXX Ἔσδρας Β΄ covers Ezra and Nehemiah together. 2ES has no Greek title because 4 Ezra survives in Latin and Syriac, not Greek.

The Synodal Esdras books are the one place the source labels mislead, so to save you the trouble:

Synodal name	label in source	is	USFM
1 Ездры	Ездры	Ezra	EZR
—	Неемии	Nehemiah	NEH
2 Ездры	1-я Ездры	1 Esdras (Greek)	1ES
3 Ездры	2-я Ездры	2 Esdras (Latin / 4 Ezra)	2ES

The source drops the "1" from canonical Ezra and then numbers the two deuterocanonical books from 1, so its "N-я Ездры" is Synodal's "N+1 Ездры". Taking those labels at face value puts 1ES and 2ES one book too early.

Licence

CC BY 4.0. Derived metrics only; no scripture text is redistributed.